Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: AW: questions using "margeff"


From   "Martin Weiss" <martin.weiss1@gmx.de>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: AW: questions using "margeff"
Date   Mon, 27 Apr 2009 23:08:54 +0200

<> 

What does " does not satisfy this property" mean? How far is it off? Too far
away for comfort? In my example, it is quite close, but that may mean
different things to different people...


*************
sysuse auto, clear

prob for we rep tr tu
margeff

di in red _b[weight]/_b[trunk]
qui mat A=e(margeff_b)
di in red A[1,1]/A[1,3]

di in red _b[weight]/_b[turn]
di in red A[1,1]/A[1,4]
*************



HTH
Martin

-----Ursprüngliche Nachricht-----
Von: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] Im Auftrag von Wei Jiang
Gesendet: Montag, 27. April 2009 16:23
An: statalist@hsphsun2.harvard.edu
Betreff: st: questions using "margeff"

I just used the "margeff" module to compute the average partial effect
(APE), and found the following problem:

Suppose a probit mode is: 

Y* = a1*x1 + a2*x2 + e;
y = 1 if Y* > 0; and = 0 if otherwise.

I would think that a1/a2 must be equal to APE(x1)/APE(x2) because 

APE(x1) = 1/n *sum(a1*normpdf(a1*x1+a2*x2))
APE(x2) = 1/n *sum(a2*normpdf(a1*x1+a2*x2))

But the APE estimates produced by margeff does not satisfy this property.
Anyone can advise?

Many thanks.

Wei Jiang



      

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index